softfloat or hardfloat ?

[Abelenda]

Hi,

out of curiosity I checked what was the default floating point
implementation for the gcc you provide for cross-compiling. [1]

It was a shock when I saw hard-float in the spec. Is it really a good
default for a processor which doesn't have FPU like the ARM920T ? I
read that in running hardfloat code on a cpu that doesn't support
floating point arithmetic is 10 times slower than using the floating
point libraries, due to the exception catch by the kernel.

So if you use a toolchain defaulting to hard implementation you need to
the option -msoft-float for every compilation.

Here are my questions : What are the options for the compilation of
packages ? Does it contain -msoft-float ? If not, is there a specific
reason ?



[1] http://wiki.openmoko.org/wiki/Toolchain